4 papers
Examining Reasoning LLMs-as-Judges in Non-Verifiable LLM Post-Training
Yixin Liu, Yue Yu, DiJia Su +7
Reasoning LLMs-as-Judges, which can benefit from inference-time scaling, provide a promising path for extending the success of reasoning models to non-verifiable domains where the…
Expert Threshold Routing for Autoregressive Language Modeling with Dynamic Computation Allocation and Load Balancing
Hanchi Sun, Yixin Liu, Yonghui Wu +1
Token-choice Mixture-of-Experts (TC-MoE) routes each token to a fixed number of experts, limiting dynamic computation allocation and requiring auxiliary losses to maintain load bal…
Evaluating Mathematical Reasoning Beyond Accuracy
Shijie Xia, Xuefeng Li, Yixin Liu +2
The leaderboard of Large Language Models (LLMs) in mathematical tasks has been continuously updated. However, the majority of evaluations focus solely on the final results, neglect…
Understanding Reference Policies in Direct Preference Optimization
Yixin Liu, Pengfei Liu, Arman Cohan
Direct Preference Optimization (DPO) has become a widely used training method for the instruction fine-tuning of large language models (LLMs). In this work, we explore an under-inv…